全自動機器翻譯加後編輯與人工翻譯之比較
Abstract
日新月異的科技演進已大幅改進機器翻譯（MT）的品質，讓機器翻譯成為從事翻譯時的輔助工具。然而，目前鮮少研究比較人工翻譯與自動機器翻譯系統加後編輯，應為一個值得研究的課題。此研究請修習翻譯課程的兩組大學生翻譯一段手機保養指南，文本為英翻中，兩組學生的語言程度具差異；兩組中一部份學生只拿到原文，另一部份學生則多了機器翻譯的譯文可進行後編輯。考量機器翻譯使用便利性及免付費，故譯文取自Google Translate。此研究計算共140位學生翻譯所花費的時間，並利用錯誤分析評估譯文準確性。統計結果顯示，有機器翻譯的譯文能夠顯著減少學生的翻譯錯誤，且使用機器翻譯能縮減學生語言程度的落差。除此之外，質性分析進一步指出學生如何使用機器翻譯譯文，以及不同組別之間的差異。此機器翻譯後端的分析提供人類輔助機器翻譯的實際應用，也點出學生翻譯上的問題。
With advances in technology, many improvements have been made to make machine translation (MT) a tool for human translators that might actually aid in the process of translation. An important issue that has not been extensively investigated is the possible efficient use of available automatic MT systems with post-editing in comparison to human translation. In this study, a cell phone care instruction guide, in need of English to Chinese translation, was given to two groups of college students of different language proficiency levels taking translation classes, where, in each group, some students received only the source text while others received the source text in addition to machine translated text for post-editing. Google Translate was the MT system used in this study for its features of free and easy accessibility. A total of 140 student subjects were timed for their translation task and the accuracy was determined by the number of errors in the translated text. The statistical results indicate the MT text is very helpful in lowering the errors in students’ translation, where the use of MT shortens the gap between students of divergent language proficiency. Further qualitative analysis elucidates how the MT text is utilized and the differences among the two groups of students. This back-end analysis of the machine translation process may offer insights into practical use of human-assisted machine translation, as well as problems encountered by students when translating.
With advances in technology, many improvements have been made to make machine translation (MT) a tool for human translators that might actually aid in the process of translation. An important issue that has not been extensively investigated is the possible efficient use of available automatic MT systems with post-editing in comparison to human translation. In this study, a cell phone care instruction guide, in need of English to Chinese translation, was given to two groups of college students of different language proficiency levels taking translation classes, where, in each group, some students received only the source text while others received the source text in addition to machine translated text for post-editing. Google Translate was the MT system used in this study for its features of free and easy accessibility. A total of 140 student subjects were timed for their translation task and the accuracy was determined by the number of errors in the translated text. The statistical results indicate the MT text is very helpful in lowering the errors in students’ translation, where the use of MT shortens the gap between students of divergent language proficiency. Further qualitative analysis elucidates how the MT text is utilized and the differences among the two groups of students. This back-end analysis of the machine translation process may offer insights into practical use of human-assisted machine translation, as well as problems encountered by students when translating.
Description
Keywords
機器翻譯, 學生翻譯, 後編輯, Google Translate, machine translation, student translation, post-editing, Google Translate